Lane Line Detection With Multi-Scale Attention and Separable Convolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current lane line detection methods in autonomous driving systems require a large number of training samples and complex neural networks, leading to low efficiency and accuracy in detecting lane lines.
Innovation Solution
A lane line detection method using a neural network model based on a multi-scale attention mechanism and deep separable convolution, combined with image post-processing techniques such as binarization, erosion, and dilation, to enhance the detection of lane lines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a complex neural network model with many training samples is used for lane line detection, then detection accuracy is improved, but detection efficiency deteriorates
Solution Approach 1:
The patent segments the convolutional operation into two separate sub-convolutional layers: a first sub-convolutional layer for spatial feature extraction and a second sub-convolutional layer for channel feature extraction. This segmentation allows the network to process features more efficiently while maintaining detection accuracy, resolving the contradiction between accuracy and efficiency.
Solution Approach 2:
The patent changes the parameters of the convolutional layers by using different kernel sizes (e.g., 3x3 for spatial, 1x1 for channel) and adjusting the number of filters at different stages. This parameter optimization enables the network to achieve high detection accuracy with reduced computational complexity, improving detection efficiency without sacrificing accuracy.
2Measurement precision
If a complex neural network model is used for lane line detection, then detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the complex convolutional operation into two simpler sub-convolutional layers with distinct functions. The first layer handles spatial relationships while the second layer handles channel relationships, reducing the overall model complexity while maintaining or improving detection accuracy through specialized feature extraction.
Solution Approach 2:
The patent applies different convolutional operations with different local qualities to different parts of the feature extraction process. The first sub-convolutional layer uses larger kernels for spatial context while the second uses 1x1 kernels for channel compression, creating locally optimized processing that reduces global model complexity.
Data Source
Figure 1
Figure 2~3
Figure 4(a)~6
AI summary
Disclosed are a lane line detection method based on deep learning, an apparatus, a storage medium, and an electronic device. The method includes: obtaining a first picture that is planned to be detected; obtaining a target feature map by inputting the first picture into a target neural network model; wherein the target neural network model includes a neural network model generated based on a multi-scale attention mechanism and a deep separable convolution model, and the target feature map is configured to represent a probability of each pixel in the first picture being a lane line pixel; and obtaining a target detection result by performing an image post-processing on the target feature map; wherein the target detection result is configured to indicate a detected lane line in the first picture.